[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-dynamo-recipe-runner":3,"mdc--5w608p-key":34,"related-org-nvidia-dynamo-recipe-runner":1739,"related-repo-nvidia-dynamo-recipe-runner":1897},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":23,"repoUrl":24,"updatedAt":25,"license":26,"forks":27,"topics":28,"repo":29,"sourceUrl":32,"mdContent":33},"dynamo-recipe-runner","deploy NVIDIA Dynamo Kubernetes recipes","Select, validate, patch, and deploy existing NVIDIA Dynamo Kubernetes recipes. Use for model\u002Fbackend\u002FGPU\u002Fdeployment-mode recipe bring-up; use router-starter for router-only mode work and troubleshoot for broken deployments.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,17,20],{"name":13,"slug":14,"type":15},"Deployment","deployment","tag",{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"AI Infrastructure","ai-infrastructure",{"name":21,"slug":22,"type":15},"Kubernetes","kubernetes",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-14T05:27:39.149297","Apache-2.0",281,[],{"repoUrl":24,"stars":23,"forks":27,"topics":30,"description":31},[],"AI agent skills published by NVIDIA","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Fdynamo-recipe-runner","---\nname: dynamo-recipe-runner\ndescription: Select, validate, patch, and deploy existing NVIDIA Dynamo Kubernetes recipes. Use for model\u002Fbackend\u002FGPU\u002Fdeployment-mode recipe bring-up; use router-starter for router-only mode work and troubleshoot for broken deployments.\nlicense: Apache-2.0\nmetadata:\n  author: Dan Gil \u003Cdagil@nvidia.com>\n  tags:\n    - dynamo\n    - kubernetes\n    - recipes\n    - bring-up\n  permissions:\n    - file_read\n    - network\n    - kubectl_exec\n---\n\n# Dynamo Recipe Runner\n\n\u003C!--\nSPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.\nSPDX-License-Identifier: CC-BY-4.0\n-->\n\n## Purpose\n\nGet from user intent to a working Dynamo recipe endpoint with minimal back and\nforth. Do not create new guide content. Operate on the existing `recipes\u002F`\ntree, patch the smallest necessary set of manifests, deploy when the user has\ncluster access, and prove success with an OpenAI-compatible smoke request.\n\n## Prerequisites\n\n- Python 3.10+ on the operator machine.\n- `kubectl` configured with a working cluster context.\n- Cluster has a default storage class for model-cache PVCs.\n- Hugging Face token stored in a Kubernetes secret named `hf-token-secret`\n  (or equivalent) in the target namespace.\n- Read access to the `recipes\u002F` tree in the ai-dynamo\u002Fdynamo repository.\n\n## Required Inputs\n\nCollect or infer these before changing manifests:\n\n- recipe target: model, framework (`vllm`, `sglang`, `trtllm`, `tokenspeed`), deployment mode, and GPU type\u002Fcount\n- Kubernetes context and namespace\n- Hugging Face secret name, usually `hf-token-secret`\n- storage class for model cache PVCs\n- runtime image tag if the recipe uses a placeholder or stale test image\n- whether to run commands or only produce exact commands\n\nIf a required value is missing and cannot be inferred from the selected recipe,\nask for only that value.\n\n## Instructions\n\n### 1. Preflight\n\nRun read-only checks first:\n\n```bash\ngit status --short\npython3 scripts\u002Frecipe_tool.py list --format table\nkubectl config current-context\nkubectl get storageclass\nkubectl get nodes -o wide\nkubectl get namespace \"${NAMESPACE}\"\nkubectl get secret hf-token-secret -n \"${NAMESPACE}\"\n```\n\nIf `kubectl` is unavailable or the cluster is unreachable, continue by\nselecting and validating the recipe, then return exact commands instead of\npretending the deployment ran.\n\n### 2. Select The Recipe\n\nUse the recipe matrix from `recipes\u002FREADME.md` and the scanner:\n\n```bash\npython3 scripts\u002Frecipe_tool.py list \\\n  --query qwen --framework vllm --mode disagg --format table\n```\n\nPrefer an exact existing recipe. Do not invent new manifests unless the user\nexplicitly asks to author a new recipe.\n\n### 3. Inspect And Validate\n\nRead the selected recipe README, model-cache manifests, `deploy.yaml`, and\n`perf.yaml` if present. Then run:\n\n```bash\npython3 scripts\u002Frecipe_tool.py validate \\\n  recipes\u002F\u003Cmodel>\u002F\u003Cframework>\u002F\u003Cmode>\n```\n\nResolve reported blockers before applying manifests: storage class, model cache\nPVC, image tag, HF token secret, GPU count, frontend service name, and router\nmode.\n\n### 4. Patch Minimal Values\n\nPatch only recipe-specific values needed for this run. Do not reformat whole\nYAML files. Common patches:\n\n- `storageClassName`\n- image repository\u002Ftag\n- model path or model cache mount path\n- GPU resource requests\u002Flimits\n- frontend `DYN_ROUTER_MODE`\n- namespace only when a manifest hardcodes it\n\nNever write Hugging Face tokens into files or logs. Use Kubernetes secrets.\n\n### 5. Deploy\n\nFollow the selected recipe README when it differs from the default sequence.\nThe default sequence is:\n\n```bash\nkubectl apply -f recipes\u002F\u003Cmodel>\u002Fmodel-cache\u002F -n \"${NAMESPACE}\"\nkubectl wait --for=condition=Complete job\u002Fmodel-download -n \"${NAMESPACE}\" --timeout=6000s\nkubectl apply -f recipes\u002F\u003Cmodel>\u002F\u003Cframework>\u002F\u003Cmode>\u002Fdeploy.yaml -n \"${NAMESPACE}\"\nkubectl get dynamographdeployment -n \"${NAMESPACE}\"\nkubectl get pods -n \"${NAMESPACE}\" -o wide\n```\n\nWait for the frontend and workers to be ready before testing.\n\n### 6. Smoke Test\n\nPort-forward the frontend service, then verify `\u002Fv1\u002Fmodels` and one chat\ncompletion:\n\n```bash\nkubectl port-forward svc\u002F\u003Cdeployment-name>-frontend 8000:8000 -n \"${NAMESPACE}\"\ncurl http:\u002F\u002F127.0.0.1:8000\u002Fv1\u002Fmodels\n```\n\nIf `dynamo-router-starter` is also installed, prefer its `scripts\u002Fcheck_router_health.py`\nfor the full OpenAI-compatible smoke test. If this fails, switch to\n`dynamo-troubleshoot`.\n\n## Available Scripts\n\n| Script | Purpose | Arguments |\n|---|---|---|\n| `scripts\u002Frecipe_tool.py list` | Enumerate available recipes, optionally filtered | `--query`, `--framework`, `--mode`, `--format` |\n| `scripts\u002Frecipe_tool.py validate` | Validate a recipe directory before apply | positional recipe path |\n\nInvoke via the agentskills.io `run_script()` protocol:\n\n```python\nrun_script(\"scripts\u002Frecipe_tool.py\", args=[\"list\", \"--framework\", \"sglang\", \"--format\", \"table\"])\nrun_script(\"scripts\u002Frecipe_tool.py\", args=[\"validate\", \"recipes\u002Fnemotron-3-super-fp8\u002Fsglang\u002Fagg\"])\n```\n\n## Examples\n\nList sglang recipes that fit a single 8xB200 node:\n\n```bash\npython3 scripts\u002Frecipe_tool.py list --framework sglang --format table\n```\n\nValidate a specific recipe and resolve blockers before applying:\n\n```bash\npython3 scripts\u002Frecipe_tool.py validate recipes\u002Fnemotron-3-super-fp8\u002Fsglang\u002Fagg\n```\n\nEquivalent through the agent protocol:\n\n```python\nrun_script(\"scripts\u002Frecipe_tool.py\", args=[\"validate\", \"recipes\u002Fnemotron-3-super-fp8\u002Fsglang\u002Fagg\"])\n```\n\n## Output Contract\n\nReturn:\n\n- selected recipe path and why it was selected\n- exact values patched\n- commands run or commands to run\n- endpoint and smoke-test result\n- unresolved blockers, if any\n- next troubleshooting step when deployment does not become healthy\n\n## Limitations\n\n- Operates on the existing `recipes\u002F` tree only. Does not author new manifests.\n- Cluster-mutating apply steps require `kubectl` permission to the target namespace.\n- Smoke-test depth is intentionally minimal; for full router\u002Fendpoint coverage use `dynamo-router-starter`.\n- Multi-node disagg transport correctness is out of scope; use `dynamo-interconnect-check` after deploy.\n\n## Troubleshooting\n\n| Symptom | Likely cause | Next step |\n|---|---|---|\n| `kubectl` cluster unreachable | Context not set or VPN down | Return exact commands instead of running them; resume when cluster is reachable |\n| `validate` reports missing storage class | Cluster has no default `StorageClass` | Patch `storageClassName` on the model-cache manifest before applying |\n| Model-cache job stuck `Pending` | PVC unbound or HF secret missing | Inspect PVC events; create or rename the HF secret to match the recipe |\n| Worker pods `ImagePullBackOff` | Stale image tag or missing pull secret | Patch the image tag; verify image pull secret in the namespace |\n| `\u002Fv1\u002Fmodels` 4xx\u002F5xx after deploy | Frontend not ready or wrong service port | Wait for pods Ready; re-run port-forward; switch to `dynamo-troubleshoot` if it persists |\n\n## Benchmark\n\nSee `BENCHMARK.md` for the NVCARPS-EVAL performance report (auto-generated by the NVSkills CI pipeline). To refresh, re-run `\u002Fnvskills-ci` on an upstream PR touching this skill.\n\n## References\n\n- Read `references\u002Fk8s-recipe-workflow.md` for command templates and readiness checks.\n- Use `scripts\u002Frecipe_tool.py` for recipe discovery and lightweight validation.\n",{"data":35,"body":46},{"name":4,"description":6,"license":26,"metadata":36},{"author":37,"tags":38,"permissions":42},"Dan Gil \u003Cdagil@nvidia.com>",[39,22,40,41],"dynamo","recipes","bring-up",[43,44,45],"file_read","network","kubectl_exec",{"type":47,"children":48},"root",[49,57,64,79,85,136,142,147,215,220,226,233,238,437,449,455,468,536,541,547,568,669,674,680,685,728,733,739,744,1022,1027,1033,1046,1126,1154,1160,1261,1274,1299,1305,1310,1349,1354,1381,1386,1399,1405,1410,1443,1449,1500,1506,1671,1677,1698,1704,1733],{"type":50,"tag":51,"props":52,"children":53},"element","h1",{"id":4},[54],{"type":55,"value":56},"text","Dynamo Recipe Runner",{"type":50,"tag":58,"props":59,"children":61},"h2",{"id":60},"purpose",[62],{"type":55,"value":63},"Purpose",{"type":50,"tag":65,"props":66,"children":67},"p",{},[68,70,77],{"type":55,"value":69},"Get from user intent to a working Dynamo recipe endpoint with minimal back and\nforth. Do not create new guide content. Operate on the existing ",{"type":50,"tag":71,"props":72,"children":74},"code",{"className":73},[],[75],{"type":55,"value":76},"recipes\u002F",{"type":55,"value":78},"\ntree, patch the smallest necessary set of manifests, deploy when the user has\ncluster access, and prove success with an OpenAI-compatible smoke request.",{"type":50,"tag":58,"props":80,"children":82},{"id":81},"prerequisites",[83],{"type":55,"value":84},"Prerequisites",{"type":50,"tag":86,"props":87,"children":88},"ul",{},[89,95,106,111,124],{"type":50,"tag":90,"props":91,"children":92},"li",{},[93],{"type":55,"value":94},"Python 3.10+ on the operator machine.",{"type":50,"tag":90,"props":96,"children":97},{},[98,104],{"type":50,"tag":71,"props":99,"children":101},{"className":100},[],[102],{"type":55,"value":103},"kubectl",{"type":55,"value":105}," configured with a working cluster context.",{"type":50,"tag":90,"props":107,"children":108},{},[109],{"type":55,"value":110},"Cluster has a default storage class for model-cache PVCs.",{"type":50,"tag":90,"props":112,"children":113},{},[114,116,122],{"type":55,"value":115},"Hugging Face token stored in a Kubernetes secret named ",{"type":50,"tag":71,"props":117,"children":119},{"className":118},[],[120],{"type":55,"value":121},"hf-token-secret",{"type":55,"value":123},"\n(or equivalent) in the target namespace.",{"type":50,"tag":90,"props":125,"children":126},{},[127,129,134],{"type":55,"value":128},"Read access to the ",{"type":50,"tag":71,"props":130,"children":132},{"className":131},[],[133],{"type":55,"value":76},{"type":55,"value":135}," tree in the ai-dynamo\u002Fdynamo repository.",{"type":50,"tag":58,"props":137,"children":139},{"id":138},"required-inputs",[140],{"type":55,"value":141},"Required Inputs",{"type":50,"tag":65,"props":143,"children":144},{},[145],{"type":55,"value":146},"Collect or infer these before changing manifests:",{"type":50,"tag":86,"props":148,"children":149},{},[150,185,190,200,205,210],{"type":50,"tag":90,"props":151,"children":152},{},[153,155,161,163,169,170,176,177,183],{"type":55,"value":154},"recipe target: model, framework (",{"type":50,"tag":71,"props":156,"children":158},{"className":157},[],[159],{"type":55,"value":160},"vllm",{"type":55,"value":162},", ",{"type":50,"tag":71,"props":164,"children":166},{"className":165},[],[167],{"type":55,"value":168},"sglang",{"type":55,"value":162},{"type":50,"tag":71,"props":171,"children":173},{"className":172},[],[174],{"type":55,"value":175},"trtllm",{"type":55,"value":162},{"type":50,"tag":71,"props":178,"children":180},{"className":179},[],[181],{"type":55,"value":182},"tokenspeed",{"type":55,"value":184},"), deployment mode, and GPU type\u002Fcount",{"type":50,"tag":90,"props":186,"children":187},{},[188],{"type":55,"value":189},"Kubernetes context and namespace",{"type":50,"tag":90,"props":191,"children":192},{},[193,195],{"type":55,"value":194},"Hugging Face secret name, usually ",{"type":50,"tag":71,"props":196,"children":198},{"className":197},[],[199],{"type":55,"value":121},{"type":50,"tag":90,"props":201,"children":202},{},[203],{"type":55,"value":204},"storage class for model cache PVCs",{"type":50,"tag":90,"props":206,"children":207},{},[208],{"type":55,"value":209},"runtime image tag if the recipe uses a placeholder or stale test image",{"type":50,"tag":90,"props":211,"children":212},{},[213],{"type":55,"value":214},"whether to run commands or only produce exact commands",{"type":50,"tag":65,"props":216,"children":217},{},[218],{"type":55,"value":219},"If a required value is missing and cannot be inferred from the selected recipe,\nask for only that value.",{"type":50,"tag":58,"props":221,"children":223},{"id":222},"instructions",[224],{"type":55,"value":225},"Instructions",{"type":50,"tag":227,"props":228,"children":230},"h3",{"id":229},"_1-preflight",[231],{"type":55,"value":232},"1. Preflight",{"type":50,"tag":65,"props":234,"children":235},{},[236],{"type":55,"value":237},"Run read-only checks first:",{"type":50,"tag":239,"props":240,"children":245},"pre",{"className":241,"code":242,"language":243,"meta":244,"style":244},"language-bash shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","git status --short\npython3 scripts\u002Frecipe_tool.py list --format table\nkubectl config current-context\nkubectl get storageclass\nkubectl get nodes -o wide\nkubectl get namespace \"${NAMESPACE}\"\nkubectl get secret hf-token-secret -n \"${NAMESPACE}\"\n","bash","",[246],{"type":50,"tag":71,"props":247,"children":248},{"__ignoreMap":244},[249,272,301,319,337,364,398],{"type":50,"tag":250,"props":251,"children":254},"span",{"class":252,"line":253},"line",1,[255,261,267],{"type":50,"tag":250,"props":256,"children":258},{"style":257},"--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B",[259],{"type":55,"value":260},"git",{"type":50,"tag":250,"props":262,"children":264},{"style":263},"--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D",[265],{"type":55,"value":266}," status",{"type":50,"tag":250,"props":268,"children":269},{"style":263},[270],{"type":55,"value":271}," --short\n",{"type":50,"tag":250,"props":273,"children":275},{"class":252,"line":274},2,[276,281,286,291,296],{"type":50,"tag":250,"props":277,"children":278},{"style":257},[279],{"type":55,"value":280},"python3",{"type":50,"tag":250,"props":282,"children":283},{"style":263},[284],{"type":55,"value":285}," scripts\u002Frecipe_tool.py",{"type":50,"tag":250,"props":287,"children":288},{"style":263},[289],{"type":55,"value":290}," list",{"type":50,"tag":250,"props":292,"children":293},{"style":263},[294],{"type":55,"value":295}," --format",{"type":50,"tag":250,"props":297,"children":298},{"style":263},[299],{"type":55,"value":300}," table\n",{"type":50,"tag":250,"props":302,"children":304},{"class":252,"line":303},3,[305,309,314],{"type":50,"tag":250,"props":306,"children":307},{"style":257},[308],{"type":55,"value":103},{"type":50,"tag":250,"props":310,"children":311},{"style":263},[312],{"type":55,"value":313}," config",{"type":50,"tag":250,"props":315,"children":316},{"style":263},[317],{"type":55,"value":318}," current-context\n",{"type":50,"tag":250,"props":320,"children":322},{"class":252,"line":321},4,[323,327,332],{"type":50,"tag":250,"props":324,"children":325},{"style":257},[326],{"type":55,"value":103},{"type":50,"tag":250,"props":328,"children":329},{"style":263},[330],{"type":55,"value":331}," get",{"type":50,"tag":250,"props":333,"children":334},{"style":263},[335],{"type":55,"value":336}," storageclass\n",{"type":50,"tag":250,"props":338,"children":340},{"class":252,"line":339},5,[341,345,349,354,359],{"type":50,"tag":250,"props":342,"children":343},{"style":257},[344],{"type":55,"value":103},{"type":50,"tag":250,"props":346,"children":347},{"style":263},[348],{"type":55,"value":331},{"type":50,"tag":250,"props":350,"children":351},{"style":263},[352],{"type":55,"value":353}," nodes",{"type":50,"tag":250,"props":355,"children":356},{"style":263},[357],{"type":55,"value":358}," -o",{"type":50,"tag":250,"props":360,"children":361},{"style":263},[362],{"type":55,"value":363}," wide\n",{"type":50,"tag":250,"props":365,"children":367},{"class":252,"line":366},6,[368,372,376,381,387,393],{"type":50,"tag":250,"props":369,"children":370},{"style":257},[371],{"type":55,"value":103},{"type":50,"tag":250,"props":373,"children":374},{"style":263},[375],{"type":55,"value":331},{"type":50,"tag":250,"props":377,"children":378},{"style":263},[379],{"type":55,"value":380}," namespace",{"type":50,"tag":250,"props":382,"children":384},{"style":383},"--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF",[385],{"type":55,"value":386}," \"${",{"type":50,"tag":250,"props":388,"children":390},{"style":389},"--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8",[391],{"type":55,"value":392},"NAMESPACE",{"type":50,"tag":250,"props":394,"children":395},{"style":383},[396],{"type":55,"value":397},"}\"\n",{"type":50,"tag":250,"props":399,"children":401},{"class":252,"line":400},7,[402,406,410,415,420,425,429,433],{"type":50,"tag":250,"props":403,"children":404},{"style":257},[405],{"type":55,"value":103},{"type":50,"tag":250,"props":407,"children":408},{"style":263},[409],{"type":55,"value":331},{"type":50,"tag":250,"props":411,"children":412},{"style":263},[413],{"type":55,"value":414}," secret",{"type":50,"tag":250,"props":416,"children":417},{"style":263},[418],{"type":55,"value":419}," hf-token-secret",{"type":50,"tag":250,"props":421,"children":422},{"style":263},[423],{"type":55,"value":424}," -n",{"type":50,"tag":250,"props":426,"children":427},{"style":383},[428],{"type":55,"value":386},{"type":50,"tag":250,"props":430,"children":431},{"style":389},[432],{"type":55,"value":392},{"type":50,"tag":250,"props":434,"children":435},{"style":383},[436],{"type":55,"value":397},{"type":50,"tag":65,"props":438,"children":439},{},[440,442,447],{"type":55,"value":441},"If ",{"type":50,"tag":71,"props":443,"children":445},{"className":444},[],[446],{"type":55,"value":103},{"type":55,"value":448}," is unavailable or the cluster is unreachable, continue by\nselecting and validating the recipe, then return exact commands instead of\npretending the deployment ran.",{"type":50,"tag":227,"props":450,"children":452},{"id":451},"_2-select-the-recipe",[453],{"type":55,"value":454},"2. Select The Recipe",{"type":50,"tag":65,"props":456,"children":457},{},[458,460,466],{"type":55,"value":459},"Use the recipe matrix from ",{"type":50,"tag":71,"props":461,"children":463},{"className":462},[],[464],{"type":55,"value":465},"recipes\u002FREADME.md",{"type":55,"value":467}," and the scanner:",{"type":50,"tag":239,"props":469,"children":471},{"className":241,"code":470,"language":243,"meta":244,"style":244},"python3 scripts\u002Frecipe_tool.py list \\\n  --query qwen --framework vllm --mode disagg --format table\n",[472],{"type":50,"tag":71,"props":473,"children":474},{"__ignoreMap":244},[475,495],{"type":50,"tag":250,"props":476,"children":477},{"class":252,"line":253},[478,482,486,490],{"type":50,"tag":250,"props":479,"children":480},{"style":257},[481],{"type":55,"value":280},{"type":50,"tag":250,"props":483,"children":484},{"style":263},[485],{"type":55,"value":285},{"type":50,"tag":250,"props":487,"children":488},{"style":263},[489],{"type":55,"value":290},{"type":50,"tag":250,"props":491,"children":492},{"style":389},[493],{"type":55,"value":494}," \\\n",{"type":50,"tag":250,"props":496,"children":497},{"class":252,"line":274},[498,503,508,513,518,523,528,532],{"type":50,"tag":250,"props":499,"children":500},{"style":263},[501],{"type":55,"value":502},"  --query",{"type":50,"tag":250,"props":504,"children":505},{"style":263},[506],{"type":55,"value":507}," qwen",{"type":50,"tag":250,"props":509,"children":510},{"style":263},[511],{"type":55,"value":512}," --framework",{"type":50,"tag":250,"props":514,"children":515},{"style":263},[516],{"type":55,"value":517}," vllm",{"type":50,"tag":250,"props":519,"children":520},{"style":263},[521],{"type":55,"value":522}," --mode",{"type":50,"tag":250,"props":524,"children":525},{"style":263},[526],{"type":55,"value":527}," disagg",{"type":50,"tag":250,"props":529,"children":530},{"style":263},[531],{"type":55,"value":295},{"type":50,"tag":250,"props":533,"children":534},{"style":263},[535],{"type":55,"value":300},{"type":50,"tag":65,"props":537,"children":538},{},[539],{"type":55,"value":540},"Prefer an exact existing recipe. Do not invent new manifests unless the user\nexplicitly asks to author a new recipe.",{"type":50,"tag":227,"props":542,"children":544},{"id":543},"_3-inspect-and-validate",[545],{"type":55,"value":546},"3. Inspect And Validate",{"type":50,"tag":65,"props":548,"children":549},{},[550,552,558,560,566],{"type":55,"value":551},"Read the selected recipe README, model-cache manifests, ",{"type":50,"tag":71,"props":553,"children":555},{"className":554},[],[556],{"type":55,"value":557},"deploy.yaml",{"type":55,"value":559},", and\n",{"type":50,"tag":71,"props":561,"children":563},{"className":562},[],[564],{"type":55,"value":565},"perf.yaml",{"type":55,"value":567}," if present. Then run:",{"type":50,"tag":239,"props":569,"children":571},{"className":241,"code":570,"language":243,"meta":244,"style":244},"python3 scripts\u002Frecipe_tool.py validate \\\n  recipes\u002F\u003Cmodel>\u002F\u003Cframework>\u002F\u003Cmode>\n",[572],{"type":50,"tag":71,"props":573,"children":574},{"__ignoreMap":244},[575,595],{"type":50,"tag":250,"props":576,"children":577},{"class":252,"line":253},[578,582,586,591],{"type":50,"tag":250,"props":579,"children":580},{"style":257},[581],{"type":55,"value":280},{"type":50,"tag":250,"props":583,"children":584},{"style":263},[585],{"type":55,"value":285},{"type":50,"tag":250,"props":587,"children":588},{"style":263},[589],{"type":55,"value":590}," validate",{"type":50,"tag":250,"props":592,"children":593},{"style":389},[594],{"type":55,"value":494},{"type":50,"tag":250,"props":596,"children":597},{"class":252,"line":274},[598,603,608,613,618,623,628,632,637,642,646,650,654,659,664],{"type":50,"tag":250,"props":599,"children":600},{"style":263},[601],{"type":55,"value":602},"  recipes\u002F",{"type":50,"tag":250,"props":604,"children":605},{"style":383},[606],{"type":55,"value":607},"\u003C",{"type":50,"tag":250,"props":609,"children":610},{"style":263},[611],{"type":55,"value":612},"mode",{"type":50,"tag":250,"props":614,"children":615},{"style":389},[616],{"type":55,"value":617},"l",{"type":50,"tag":250,"props":619,"children":620},{"style":383},[621],{"type":55,"value":622},">",{"type":50,"tag":250,"props":624,"children":625},{"style":263},[626],{"type":55,"value":627},"\u002F",{"type":50,"tag":250,"props":629,"children":630},{"style":383},[631],{"type":55,"value":607},{"type":50,"tag":250,"props":633,"children":634},{"style":263},[635],{"type":55,"value":636},"framewor",{"type":50,"tag":250,"props":638,"children":639},{"style":389},[640],{"type":55,"value":641},"k",{"type":50,"tag":250,"props":643,"children":644},{"style":383},[645],{"type":55,"value":622},{"type":50,"tag":250,"props":647,"children":648},{"style":263},[649],{"type":55,"value":627},{"type":50,"tag":250,"props":651,"children":652},{"style":383},[653],{"type":55,"value":607},{"type":50,"tag":250,"props":655,"children":656},{"style":263},[657],{"type":55,"value":658},"mod",{"type":50,"tag":250,"props":660,"children":661},{"style":389},[662],{"type":55,"value":663},"e",{"type":50,"tag":250,"props":665,"children":666},{"style":383},[667],{"type":55,"value":668},">\n",{"type":50,"tag":65,"props":670,"children":671},{},[672],{"type":55,"value":673},"Resolve reported blockers before applying manifests: storage class, model cache\nPVC, image tag, HF token secret, GPU count, frontend service name, and router\nmode.",{"type":50,"tag":227,"props":675,"children":677},{"id":676},"_4-patch-minimal-values",[678],{"type":55,"value":679},"4. Patch Minimal Values",{"type":50,"tag":65,"props":681,"children":682},{},[683],{"type":55,"value":684},"Patch only recipe-specific values needed for this run. Do not reformat whole\nYAML files. Common patches:",{"type":50,"tag":86,"props":686,"children":687},{},[688,697,702,707,712,723],{"type":50,"tag":90,"props":689,"children":690},{},[691],{"type":50,"tag":71,"props":692,"children":694},{"className":693},[],[695],{"type":55,"value":696},"storageClassName",{"type":50,"tag":90,"props":698,"children":699},{},[700],{"type":55,"value":701},"image repository\u002Ftag",{"type":50,"tag":90,"props":703,"children":704},{},[705],{"type":55,"value":706},"model path or model cache mount path",{"type":50,"tag":90,"props":708,"children":709},{},[710],{"type":55,"value":711},"GPU resource requests\u002Flimits",{"type":50,"tag":90,"props":713,"children":714},{},[715,717],{"type":55,"value":716},"frontend ",{"type":50,"tag":71,"props":718,"children":720},{"className":719},[],[721],{"type":55,"value":722},"DYN_ROUTER_MODE",{"type":50,"tag":90,"props":724,"children":725},{},[726],{"type":55,"value":727},"namespace only when a manifest hardcodes it",{"type":50,"tag":65,"props":729,"children":730},{},[731],{"type":55,"value":732},"Never write Hugging Face tokens into files or logs. Use Kubernetes secrets.",{"type":50,"tag":227,"props":734,"children":736},{"id":735},"_5-deploy",[737],{"type":55,"value":738},"5. Deploy",{"type":50,"tag":65,"props":740,"children":741},{},[742],{"type":55,"value":743},"Follow the selected recipe README when it differs from the default sequence.\nThe default sequence is:",{"type":50,"tag":239,"props":745,"children":747},{"className":241,"code":746,"language":243,"meta":244,"style":244},"kubectl apply -f recipes\u002F\u003Cmodel>\u002Fmodel-cache\u002F -n \"${NAMESPACE}\"\nkubectl wait --for=condition=Complete job\u002Fmodel-download -n \"${NAMESPACE}\" --timeout=6000s\nkubectl apply -f recipes\u002F\u003Cmodel>\u002F\u003Cframework>\u002F\u003Cmode>\u002Fdeploy.yaml -n \"${NAMESPACE}\"\nkubectl get dynamographdeployment -n \"${NAMESPACE}\"\nkubectl get pods -n \"${NAMESPACE}\" -o wide\n",[748],{"type":50,"tag":71,"props":749,"children":750},{"__ignoreMap":244},[751,810,854,950,982],{"type":50,"tag":250,"props":752,"children":753},{"class":252,"line":253},[754,758,763,768,773,777,781,785,789,794,798,802,806],{"type":50,"tag":250,"props":755,"children":756},{"style":257},[757],{"type":55,"value":103},{"type":50,"tag":250,"props":759,"children":760},{"style":263},[761],{"type":55,"value":762}," apply",{"type":50,"tag":250,"props":764,"children":765},{"style":263},[766],{"type":55,"value":767}," -f",{"type":50,"tag":250,"props":769,"children":770},{"style":263},[771],{"type":55,"value":772}," recipes\u002F",{"type":50,"tag":250,"props":774,"children":775},{"style":383},[776],{"type":55,"value":607},{"type":50,"tag":250,"props":778,"children":779},{"style":263},[780],{"type":55,"value":612},{"type":50,"tag":250,"props":782,"children":783},{"style":389},[784],{"type":55,"value":617},{"type":50,"tag":250,"props":786,"children":787},{"style":383},[788],{"type":55,"value":622},{"type":50,"tag":250,"props":790,"children":791},{"style":263},[792],{"type":55,"value":793},"\u002Fmodel-cache\u002F",{"type":50,"tag":250,"props":795,"children":796},{"style":263},[797],{"type":55,"value":424},{"type":50,"tag":250,"props":799,"children":800},{"style":383},[801],{"type":55,"value":386},{"type":50,"tag":250,"props":803,"children":804},{"style":389},[805],{"type":55,"value":392},{"type":50,"tag":250,"props":807,"children":808},{"style":383},[809],{"type":55,"value":397},{"type":50,"tag":250,"props":811,"children":812},{"class":252,"line":274},[813,817,822,827,832,836,840,844,849],{"type":50,"tag":250,"props":814,"children":815},{"style":257},[816],{"type":55,"value":103},{"type":50,"tag":250,"props":818,"children":819},{"style":263},[820],{"type":55,"value":821}," wait",{"type":50,"tag":250,"props":823,"children":824},{"style":263},[825],{"type":55,"value":826}," --for=condition=Complete",{"type":50,"tag":250,"props":828,"children":829},{"style":263},[830],{"type":55,"value":831}," job\u002Fmodel-download",{"type":50,"tag":250,"props":833,"children":834},{"style":263},[835],{"type":55,"value":424},{"type":50,"tag":250,"props":837,"children":838},{"style":383},[839],{"type":55,"value":386},{"type":50,"tag":250,"props":841,"children":842},{"style":389},[843],{"type":55,"value":392},{"type":50,"tag":250,"props":845,"children":846},{"style":383},[847],{"type":55,"value":848},"}\"",{"type":50,"tag":250,"props":850,"children":851},{"style":263},[852],{"type":55,"value":853}," --timeout=6000s\n",{"type":50,"tag":250,"props":855,"children":856},{"class":252,"line":303},[857,861,865,869,873,877,881,885,889,893,897,901,905,909,913,917,921,925,929,934,938,942,946],{"type":50,"tag":250,"props":858,"children":859},{"style":257},[860],{"type":55,"value":103},{"type":50,"tag":250,"props":862,"children":863},{"style":263},[864],{"type":55,"value":762},{"type":50,"tag":250,"props":866,"children":867},{"style":263},[868],{"type":55,"value":767},{"type":50,"tag":250,"props":870,"children":871},{"style":263},[872],{"type":55,"value":772},{"type":50,"tag":250,"props":874,"children":875},{"style":383},[876],{"type":55,"value":607},{"type":50,"tag":250,"props":878,"children":879},{"style":263},[880],{"type":55,"value":612},{"type":50,"tag":250,"props":882,"children":883},{"style":389},[884],{"type":55,"value":617},{"type":50,"tag":250,"props":886,"children":887},{"style":383},[888],{"type":55,"value":622},{"type":50,"tag":250,"props":890,"children":891},{"style":263},[892],{"type":55,"value":627},{"type":50,"tag":250,"props":894,"children":895},{"style":383},[896],{"type":55,"value":607},{"type":50,"tag":250,"props":898,"children":899},{"style":263},[900],{"type":55,"value":636},{"type":50,"tag":250,"props":902,"children":903},{"style":389},[904],{"type":55,"value":641},{"type":50,"tag":250,"props":906,"children":907},{"style":383},[908],{"type":55,"value":622},{"type":50,"tag":250,"props":910,"children":911},{"style":263},[912],{"type":55,"value":627},{"type":50,"tag":250,"props":914,"children":915},{"style":383},[916],{"type":55,"value":607},{"type":50,"tag":250,"props":918,"children":919},{"style":263},[920],{"type":55,"value":658},{"type":50,"tag":250,"props":922,"children":923},{"style":389},[924],{"type":55,"value":663},{"type":50,"tag":250,"props":926,"children":927},{"style":383},[928],{"type":55,"value":622},{"type":50,"tag":250,"props":930,"children":931},{"style":263},[932],{"type":55,"value":933},"\u002Fdeploy.yaml",{"type":50,"tag":250,"props":935,"children":936},{"style":263},[937],{"type":55,"value":424},{"type":50,"tag":250,"props":939,"children":940},{"style":383},[941],{"type":55,"value":386},{"type":50,"tag":250,"props":943,"children":944},{"style":389},[945],{"type":55,"value":392},{"type":50,"tag":250,"props":947,"children":948},{"style":383},[949],{"type":55,"value":397},{"type":50,"tag":250,"props":951,"children":952},{"class":252,"line":321},[953,957,961,966,970,974,978],{"type":50,"tag":250,"props":954,"children":955},{"style":257},[956],{"type":55,"value":103},{"type":50,"tag":250,"props":958,"children":959},{"style":263},[960],{"type":55,"value":331},{"type":50,"tag":250,"props":962,"children":963},{"style":263},[964],{"type":55,"value":965}," dynamographdeployment",{"type":50,"tag":250,"props":967,"children":968},{"style":263},[969],{"type":55,"value":424},{"type":50,"tag":250,"props":971,"children":972},{"style":383},[973],{"type":55,"value":386},{"type":50,"tag":250,"props":975,"children":976},{"style":389},[977],{"type":55,"value":392},{"type":50,"tag":250,"props":979,"children":980},{"style":383},[981],{"type":55,"value":397},{"type":50,"tag":250,"props":983,"children":984},{"class":252,"line":339},[985,989,993,998,1002,1006,1010,1014,1018],{"type":50,"tag":250,"props":986,"children":987},{"style":257},[988],{"type":55,"value":103},{"type":50,"tag":250,"props":990,"children":991},{"style":263},[992],{"type":55,"value":331},{"type":50,"tag":250,"props":994,"children":995},{"style":263},[996],{"type":55,"value":997}," pods",{"type":50,"tag":250,"props":999,"children":1000},{"style":263},[1001],{"type":55,"value":424},{"type":50,"tag":250,"props":1003,"children":1004},{"style":383},[1005],{"type":55,"value":386},{"type":50,"tag":250,"props":1007,"children":1008},{"style":389},[1009],{"type":55,"value":392},{"type":50,"tag":250,"props":1011,"children":1012},{"style":383},[1013],{"type":55,"value":848},{"type":50,"tag":250,"props":1015,"children":1016},{"style":263},[1017],{"type":55,"value":358},{"type":50,"tag":250,"props":1019,"children":1020},{"style":263},[1021],{"type":55,"value":363},{"type":50,"tag":65,"props":1023,"children":1024},{},[1025],{"type":55,"value":1026},"Wait for the frontend and workers to be ready before testing.",{"type":50,"tag":227,"props":1028,"children":1030},{"id":1029},"_6-smoke-test",[1031],{"type":55,"value":1032},"6. Smoke Test",{"type":50,"tag":65,"props":1034,"children":1035},{},[1036,1038,1044],{"type":55,"value":1037},"Port-forward the frontend service, then verify ",{"type":50,"tag":71,"props":1039,"children":1041},{"className":1040},[],[1042],{"type":55,"value":1043},"\u002Fv1\u002Fmodels",{"type":55,"value":1045}," and one chat\ncompletion:",{"type":50,"tag":239,"props":1047,"children":1049},{"className":241,"code":1048,"language":243,"meta":244,"style":244},"kubectl port-forward svc\u002F\u003Cdeployment-name>-frontend 8000:8000 -n \"${NAMESPACE}\"\ncurl http:\u002F\u002F127.0.0.1:8000\u002Fv1\u002Fmodels\n",[1050],{"type":50,"tag":71,"props":1051,"children":1052},{"__ignoreMap":244},[1053,1113],{"type":50,"tag":250,"props":1054,"children":1055},{"class":252,"line":253},[1056,1060,1065,1070,1074,1079,1083,1087,1092,1097,1101,1105,1109],{"type":50,"tag":250,"props":1057,"children":1058},{"style":257},[1059],{"type":55,"value":103},{"type":50,"tag":250,"props":1061,"children":1062},{"style":263},[1063],{"type":55,"value":1064}," port-forward",{"type":50,"tag":250,"props":1066,"children":1067},{"style":263},[1068],{"type":55,"value":1069}," svc\u002F",{"type":50,"tag":250,"props":1071,"children":1072},{"style":383},[1073],{"type":55,"value":607},{"type":50,"tag":250,"props":1075,"children":1076},{"style":263},[1077],{"type":55,"value":1078},"deployment-nam",{"type":50,"tag":250,"props":1080,"children":1081},{"style":389},[1082],{"type":55,"value":663},{"type":50,"tag":250,"props":1084,"children":1085},{"style":383},[1086],{"type":55,"value":622},{"type":50,"tag":250,"props":1088,"children":1089},{"style":263},[1090],{"type":55,"value":1091},"-frontend",{"type":50,"tag":250,"props":1093,"children":1094},{"style":263},[1095],{"type":55,"value":1096}," 8000:8000",{"type":50,"tag":250,"props":1098,"children":1099},{"style":263},[1100],{"type":55,"value":424},{"type":50,"tag":250,"props":1102,"children":1103},{"style":383},[1104],{"type":55,"value":386},{"type":50,"tag":250,"props":1106,"children":1107},{"style":389},[1108],{"type":55,"value":392},{"type":50,"tag":250,"props":1110,"children":1111},{"style":383},[1112],{"type":55,"value":397},{"type":50,"tag":250,"props":1114,"children":1115},{"class":252,"line":274},[1116,1121],{"type":50,"tag":250,"props":1117,"children":1118},{"style":257},[1119],{"type":55,"value":1120},"curl",{"type":50,"tag":250,"props":1122,"children":1123},{"style":263},[1124],{"type":55,"value":1125}," http:\u002F\u002F127.0.0.1:8000\u002Fv1\u002Fmodels\n",{"type":50,"tag":65,"props":1127,"children":1128},{},[1129,1130,1136,1138,1144,1146,1152],{"type":55,"value":441},{"type":50,"tag":71,"props":1131,"children":1133},{"className":1132},[],[1134],{"type":55,"value":1135},"dynamo-router-starter",{"type":55,"value":1137}," is also installed, prefer its ",{"type":50,"tag":71,"props":1139,"children":1141},{"className":1140},[],[1142],{"type":55,"value":1143},"scripts\u002Fcheck_router_health.py",{"type":55,"value":1145},"\nfor the full OpenAI-compatible smoke test. If this fails, switch to\n",{"type":50,"tag":71,"props":1147,"children":1149},{"className":1148},[],[1150],{"type":55,"value":1151},"dynamo-troubleshoot",{"type":55,"value":1153},".",{"type":50,"tag":58,"props":1155,"children":1157},{"id":1156},"available-scripts",[1158],{"type":55,"value":1159},"Available Scripts",{"type":50,"tag":1161,"props":1162,"children":1163},"table",{},[1164,1187],{"type":50,"tag":1165,"props":1166,"children":1167},"thead",{},[1168],{"type":50,"tag":1169,"props":1170,"children":1171},"tr",{},[1172,1178,1182],{"type":50,"tag":1173,"props":1174,"children":1175},"th",{},[1176],{"type":55,"value":1177},"Script",{"type":50,"tag":1173,"props":1179,"children":1180},{},[1181],{"type":55,"value":63},{"type":50,"tag":1173,"props":1183,"children":1184},{},[1185],{"type":55,"value":1186},"Arguments",{"type":50,"tag":1188,"props":1189,"children":1190},"tbody",{},[1191,1239],{"type":50,"tag":1169,"props":1192,"children":1193},{},[1194,1204,1209],{"type":50,"tag":1195,"props":1196,"children":1197},"td",{},[1198],{"type":50,"tag":71,"props":1199,"children":1201},{"className":1200},[],[1202],{"type":55,"value":1203},"scripts\u002Frecipe_tool.py list",{"type":50,"tag":1195,"props":1205,"children":1206},{},[1207],{"type":55,"value":1208},"Enumerate available recipes, optionally filtered",{"type":50,"tag":1195,"props":1210,"children":1211},{},[1212,1218,1219,1225,1226,1232,1233],{"type":50,"tag":71,"props":1213,"children":1215},{"className":1214},[],[1216],{"type":55,"value":1217},"--query",{"type":55,"value":162},{"type":50,"tag":71,"props":1220,"children":1222},{"className":1221},[],[1223],{"type":55,"value":1224},"--framework",{"type":55,"value":162},{"type":50,"tag":71,"props":1227,"children":1229},{"className":1228},[],[1230],{"type":55,"value":1231},"--mode",{"type":55,"value":162},{"type":50,"tag":71,"props":1234,"children":1236},{"className":1235},[],[1237],{"type":55,"value":1238},"--format",{"type":50,"tag":1169,"props":1240,"children":1241},{},[1242,1251,1256],{"type":50,"tag":1195,"props":1243,"children":1244},{},[1245],{"type":50,"tag":71,"props":1246,"children":1248},{"className":1247},[],[1249],{"type":55,"value":1250},"scripts\u002Frecipe_tool.py validate",{"type":50,"tag":1195,"props":1252,"children":1253},{},[1254],{"type":55,"value":1255},"Validate a recipe directory before apply",{"type":50,"tag":1195,"props":1257,"children":1258},{},[1259],{"type":55,"value":1260},"positional recipe path",{"type":50,"tag":65,"props":1262,"children":1263},{},[1264,1266,1272],{"type":55,"value":1265},"Invoke via the agentskills.io ",{"type":50,"tag":71,"props":1267,"children":1269},{"className":1268},[],[1270],{"type":55,"value":1271},"run_script()",{"type":55,"value":1273}," protocol:",{"type":50,"tag":239,"props":1275,"children":1279},{"className":1276,"code":1277,"language":1278,"meta":244,"style":244},"language-python shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","run_script(\"scripts\u002Frecipe_tool.py\", args=[\"list\", \"--framework\", \"sglang\", \"--format\", \"table\"])\nrun_script(\"scripts\u002Frecipe_tool.py\", args=[\"validate\", \"recipes\u002Fnemotron-3-super-fp8\u002Fsglang\u002Fagg\"])\n","python",[1280],{"type":50,"tag":71,"props":1281,"children":1282},{"__ignoreMap":244},[1283,1291],{"type":50,"tag":250,"props":1284,"children":1285},{"class":252,"line":253},[1286],{"type":50,"tag":250,"props":1287,"children":1288},{},[1289],{"type":55,"value":1290},"run_script(\"scripts\u002Frecipe_tool.py\", args=[\"list\", \"--framework\", \"sglang\", \"--format\", \"table\"])\n",{"type":50,"tag":250,"props":1292,"children":1293},{"class":252,"line":274},[1294],{"type":50,"tag":250,"props":1295,"children":1296},{},[1297],{"type":55,"value":1298},"run_script(\"scripts\u002Frecipe_tool.py\", args=[\"validate\", \"recipes\u002Fnemotron-3-super-fp8\u002Fsglang\u002Fagg\"])\n",{"type":50,"tag":58,"props":1300,"children":1302},{"id":1301},"examples",[1303],{"type":55,"value":1304},"Examples",{"type":50,"tag":65,"props":1306,"children":1307},{},[1308],{"type":55,"value":1309},"List sglang recipes that fit a single 8xB200 node:",{"type":50,"tag":239,"props":1311,"children":1313},{"className":241,"code":1312,"language":243,"meta":244,"style":244},"python3 scripts\u002Frecipe_tool.py list --framework sglang --format table\n",[1314],{"type":50,"tag":71,"props":1315,"children":1316},{"__ignoreMap":244},[1317],{"type":50,"tag":250,"props":1318,"children":1319},{"class":252,"line":253},[1320,1324,1328,1332,1336,1341,1345],{"type":50,"tag":250,"props":1321,"children":1322},{"style":257},[1323],{"type":55,"value":280},{"type":50,"tag":250,"props":1325,"children":1326},{"style":263},[1327],{"type":55,"value":285},{"type":50,"tag":250,"props":1329,"children":1330},{"style":263},[1331],{"type":55,"value":290},{"type":50,"tag":250,"props":1333,"children":1334},{"style":263},[1335],{"type":55,"value":512},{"type":50,"tag":250,"props":1337,"children":1338},{"style":263},[1339],{"type":55,"value":1340}," sglang",{"type":50,"tag":250,"props":1342,"children":1343},{"style":263},[1344],{"type":55,"value":295},{"type":50,"tag":250,"props":1346,"children":1347},{"style":263},[1348],{"type":55,"value":300},{"type":50,"tag":65,"props":1350,"children":1351},{},[1352],{"type":55,"value":1353},"Validate a specific recipe and resolve blockers before applying:",{"type":50,"tag":239,"props":1355,"children":1357},{"className":241,"code":1356,"language":243,"meta":244,"style":244},"python3 scripts\u002Frecipe_tool.py validate recipes\u002Fnemotron-3-super-fp8\u002Fsglang\u002Fagg\n",[1358],{"type":50,"tag":71,"props":1359,"children":1360},{"__ignoreMap":244},[1361],{"type":50,"tag":250,"props":1362,"children":1363},{"class":252,"line":253},[1364,1368,1372,1376],{"type":50,"tag":250,"props":1365,"children":1366},{"style":257},[1367],{"type":55,"value":280},{"type":50,"tag":250,"props":1369,"children":1370},{"style":263},[1371],{"type":55,"value":285},{"type":50,"tag":250,"props":1373,"children":1374},{"style":263},[1375],{"type":55,"value":590},{"type":50,"tag":250,"props":1377,"children":1378},{"style":263},[1379],{"type":55,"value":1380}," recipes\u002Fnemotron-3-super-fp8\u002Fsglang\u002Fagg\n",{"type":50,"tag":65,"props":1382,"children":1383},{},[1384],{"type":55,"value":1385},"Equivalent through the agent protocol:",{"type":50,"tag":239,"props":1387,"children":1388},{"className":1276,"code":1298,"language":1278,"meta":244,"style":244},[1389],{"type":50,"tag":71,"props":1390,"children":1391},{"__ignoreMap":244},[1392],{"type":50,"tag":250,"props":1393,"children":1394},{"class":252,"line":253},[1395],{"type":50,"tag":250,"props":1396,"children":1397},{},[1398],{"type":55,"value":1298},{"type":50,"tag":58,"props":1400,"children":1402},{"id":1401},"output-contract",[1403],{"type":55,"value":1404},"Output Contract",{"type":50,"tag":65,"props":1406,"children":1407},{},[1408],{"type":55,"value":1409},"Return:",{"type":50,"tag":86,"props":1411,"children":1412},{},[1413,1418,1423,1428,1433,1438],{"type":50,"tag":90,"props":1414,"children":1415},{},[1416],{"type":55,"value":1417},"selected recipe path and why it was selected",{"type":50,"tag":90,"props":1419,"children":1420},{},[1421],{"type":55,"value":1422},"exact values patched",{"type":50,"tag":90,"props":1424,"children":1425},{},[1426],{"type":55,"value":1427},"commands run or commands to run",{"type":50,"tag":90,"props":1429,"children":1430},{},[1431],{"type":55,"value":1432},"endpoint and smoke-test result",{"type":50,"tag":90,"props":1434,"children":1435},{},[1436],{"type":55,"value":1437},"unresolved blockers, if any",{"type":50,"tag":90,"props":1439,"children":1440},{},[1441],{"type":55,"value":1442},"next troubleshooting step when deployment does not become healthy",{"type":50,"tag":58,"props":1444,"children":1446},{"id":1445},"limitations",[1447],{"type":55,"value":1448},"Limitations",{"type":50,"tag":86,"props":1450,"children":1451},{},[1452,1464,1476,1487],{"type":50,"tag":90,"props":1453,"children":1454},{},[1455,1457,1462],{"type":55,"value":1456},"Operates on the existing ",{"type":50,"tag":71,"props":1458,"children":1460},{"className":1459},[],[1461],{"type":55,"value":76},{"type":55,"value":1463}," tree only. Does not author new manifests.",{"type":50,"tag":90,"props":1465,"children":1466},{},[1467,1469,1474],{"type":55,"value":1468},"Cluster-mutating apply steps require ",{"type":50,"tag":71,"props":1470,"children":1472},{"className":1471},[],[1473],{"type":55,"value":103},{"type":55,"value":1475}," permission to the target namespace.",{"type":50,"tag":90,"props":1477,"children":1478},{},[1479,1481,1486],{"type":55,"value":1480},"Smoke-test depth is intentionally minimal; for full router\u002Fendpoint coverage use ",{"type":50,"tag":71,"props":1482,"children":1484},{"className":1483},[],[1485],{"type":55,"value":1135},{"type":55,"value":1153},{"type":50,"tag":90,"props":1488,"children":1489},{},[1490,1492,1498],{"type":55,"value":1491},"Multi-node disagg transport correctness is out of scope; use ",{"type":50,"tag":71,"props":1493,"children":1495},{"className":1494},[],[1496],{"type":55,"value":1497},"dynamo-interconnect-check",{"type":55,"value":1499}," after deploy.",{"type":50,"tag":58,"props":1501,"children":1503},{"id":1502},"troubleshooting",[1504],{"type":55,"value":1505},"Troubleshooting",{"type":50,"tag":1161,"props":1507,"children":1508},{},[1509,1530],{"type":50,"tag":1165,"props":1510,"children":1511},{},[1512],{"type":50,"tag":1169,"props":1513,"children":1514},{},[1515,1520,1525],{"type":50,"tag":1173,"props":1516,"children":1517},{},[1518],{"type":55,"value":1519},"Symptom",{"type":50,"tag":1173,"props":1521,"children":1522},{},[1523],{"type":55,"value":1524},"Likely cause",{"type":50,"tag":1173,"props":1526,"children":1527},{},[1528],{"type":55,"value":1529},"Next step",{"type":50,"tag":1188,"props":1531,"children":1532},{},[1533,1556,1593,1617,1641],{"type":50,"tag":1169,"props":1534,"children":1535},{},[1536,1546,1551],{"type":50,"tag":1195,"props":1537,"children":1538},{},[1539,1544],{"type":50,"tag":71,"props":1540,"children":1542},{"className":1541},[],[1543],{"type":55,"value":103},{"type":55,"value":1545}," cluster unreachable",{"type":50,"tag":1195,"props":1547,"children":1548},{},[1549],{"type":55,"value":1550},"Context not set or VPN down",{"type":50,"tag":1195,"props":1552,"children":1553},{},[1554],{"type":55,"value":1555},"Return exact commands instead of running them; resume when cluster is reachable",{"type":50,"tag":1169,"props":1557,"children":1558},{},[1559,1570,1581],{"type":50,"tag":1195,"props":1560,"children":1561},{},[1562,1568],{"type":50,"tag":71,"props":1563,"children":1565},{"className":1564},[],[1566],{"type":55,"value":1567},"validate",{"type":55,"value":1569}," reports missing storage class",{"type":50,"tag":1195,"props":1571,"children":1572},{},[1573,1575],{"type":55,"value":1574},"Cluster has no default ",{"type":50,"tag":71,"props":1576,"children":1578},{"className":1577},[],[1579],{"type":55,"value":1580},"StorageClass",{"type":50,"tag":1195,"props":1582,"children":1583},{},[1584,1586,1591],{"type":55,"value":1585},"Patch ",{"type":50,"tag":71,"props":1587,"children":1589},{"className":1588},[],[1590],{"type":55,"value":696},{"type":55,"value":1592}," on the model-cache manifest before applying",{"type":50,"tag":1169,"props":1594,"children":1595},{},[1596,1607,1612],{"type":50,"tag":1195,"props":1597,"children":1598},{},[1599,1601],{"type":55,"value":1600},"Model-cache job stuck ",{"type":50,"tag":71,"props":1602,"children":1604},{"className":1603},[],[1605],{"type":55,"value":1606},"Pending",{"type":50,"tag":1195,"props":1608,"children":1609},{},[1610],{"type":55,"value":1611},"PVC unbound or HF secret missing",{"type":50,"tag":1195,"props":1613,"children":1614},{},[1615],{"type":55,"value":1616},"Inspect PVC events; create or rename the HF secret to match the recipe",{"type":50,"tag":1169,"props":1618,"children":1619},{},[1620,1631,1636],{"type":50,"tag":1195,"props":1621,"children":1622},{},[1623,1625],{"type":55,"value":1624},"Worker pods ",{"type":50,"tag":71,"props":1626,"children":1628},{"className":1627},[],[1629],{"type":55,"value":1630},"ImagePullBackOff",{"type":50,"tag":1195,"props":1632,"children":1633},{},[1634],{"type":55,"value":1635},"Stale image tag or missing pull secret",{"type":50,"tag":1195,"props":1637,"children":1638},{},[1639],{"type":55,"value":1640},"Patch the image tag; verify image pull secret in the namespace",{"type":50,"tag":1169,"props":1642,"children":1643},{},[1644,1654,1659],{"type":50,"tag":1195,"props":1645,"children":1646},{},[1647,1652],{"type":50,"tag":71,"props":1648,"children":1650},{"className":1649},[],[1651],{"type":55,"value":1043},{"type":55,"value":1653}," 4xx\u002F5xx after deploy",{"type":50,"tag":1195,"props":1655,"children":1656},{},[1657],{"type":55,"value":1658},"Frontend not ready or wrong service port",{"type":50,"tag":1195,"props":1660,"children":1661},{},[1662,1664,1669],{"type":55,"value":1663},"Wait for pods Ready; re-run port-forward; switch to ",{"type":50,"tag":71,"props":1665,"children":1667},{"className":1666},[],[1668],{"type":55,"value":1151},{"type":55,"value":1670}," if it persists",{"type":50,"tag":58,"props":1672,"children":1674},{"id":1673},"benchmark",[1675],{"type":55,"value":1676},"Benchmark",{"type":50,"tag":65,"props":1678,"children":1679},{},[1680,1682,1688,1690,1696],{"type":55,"value":1681},"See ",{"type":50,"tag":71,"props":1683,"children":1685},{"className":1684},[],[1686],{"type":55,"value":1687},"BENCHMARK.md",{"type":55,"value":1689}," for the NVCARPS-EVAL performance report (auto-generated by the NVSkills CI pipeline). To refresh, re-run ",{"type":50,"tag":71,"props":1691,"children":1693},{"className":1692},[],[1694],{"type":55,"value":1695},"\u002Fnvskills-ci",{"type":55,"value":1697}," on an upstream PR touching this skill.",{"type":50,"tag":58,"props":1699,"children":1701},{"id":1700},"references",[1702],{"type":55,"value":1703},"References",{"type":50,"tag":86,"props":1705,"children":1706},{},[1707,1720],{"type":50,"tag":90,"props":1708,"children":1709},{},[1710,1712,1718],{"type":55,"value":1711},"Read ",{"type":50,"tag":71,"props":1713,"children":1715},{"className":1714},[],[1716],{"type":55,"value":1717},"references\u002Fk8s-recipe-workflow.md",{"type":55,"value":1719}," for command templates and readiness checks.",{"type":50,"tag":90,"props":1721,"children":1722},{},[1723,1725,1731],{"type":55,"value":1724},"Use ",{"type":50,"tag":71,"props":1726,"children":1728},{"className":1727},[],[1729],{"type":55,"value":1730},"scripts\u002Frecipe_tool.py",{"type":55,"value":1732}," for recipe discovery and lightweight validation.",{"type":50,"tag":1734,"props":1735,"children":1736},"style",{},[1737],{"type":55,"value":1738},"html .light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html.light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}",{"items":1740,"total":1896},[1741,1759,1774,1785,1797,1811,1824,1838,1851,1862,1876,1885],{"slug":1742,"name":1742,"fn":1743,"description":1744,"org":1745,"tags":1746,"stars":1756,"repoUrl":1757,"updatedAt":1758},"nemoclaw-user-guide","retrieve NemoClaw documentation and configuration","Guides human users' AI agents to the NemoClaw docs MCP server and canonical Fern documentation in Markdown form. Use when users ask how to install, configure, operate, troubleshoot, secure, or learn NemoClaw with an AI coding assistant. Trigger keywords - nemoclaw docs, use nemoclaw with ai agent, nemoclaw mcp docs, nemoclaw install help, nemoclaw quickstart, nemoclaw markdown docs, llms.txt, agent skills.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1747,1750,1753],{"name":1748,"slug":1749,"type":15},"Documentation","documentation",{"name":1751,"slug":1752,"type":15},"MCP","mcp",{"name":1754,"slug":1755,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":1760,"name":1760,"fn":1761,"description":1762,"org":1763,"tags":1764,"stars":1771,"repoUrl":1772,"updatedAt":1773},"mcore-build-and-dependency","manage Megatron-LM development environments","Container-based dev environment setup and dependency management for Megatron-LM. Covers acquiring and launching the CI container, uv package management, and updating uv.lock.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1765,1768,1769],{"name":1766,"slug":1767,"type":15},"Containers","containers",{"name":13,"slug":14,"type":15},{"name":1770,"slug":1278,"type":15},"Python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":1775,"name":1775,"fn":1776,"description":1777,"org":1778,"tags":1779,"stars":1771,"repoUrl":1772,"updatedAt":1784},"mcore-bump-base-image","update NVIDIA PyTorch base images","Bump the NVIDIA PyTorch base image (`nvcr.io\u002Fnvidia\u002Fpytorch:YY.MM-py3`) used by Megatron-LM CI. Covers the two pin sites (GitHub CI in `docker\u002F.ngc_version.dev` and GitLab CI in `.gitlab\u002Fstages\u002F01.build.yml`), the post-bump CI loop (re-run functional tests, refresh golden values, mark broken tests), and the gotchas that bit PRs",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1780,1783],{"name":1781,"slug":1782,"type":15},"CI\u002FCD","ci-cd",{"name":13,"slug":14,"type":15},"2026-07-14T05:25:59.97109",{"slug":1786,"name":1786,"fn":1787,"description":1788,"org":1789,"tags":1790,"stars":1771,"repoUrl":1772,"updatedAt":1796},"mcore-cicd","manage CI\u002FCD pipelines for Megatron-LM","CI\u002FCD reference for Megatron-LM. Covers CI pipeline structure, PR scope labels, triggering internal GitLab CI (which force-pushes the current branch to a pull-request\u002FBRANCH ref — always dry-run and verify the destination first; never run against shared or protected branches), and CI failure investigation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1791,1792,1793],{"name":1781,"slug":1782,"type":15},{"name":13,"slug":14,"type":15},{"name":1794,"slug":1795,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":1798,"name":1798,"fn":1799,"description":1800,"org":1801,"tags":1802,"stars":1771,"repoUrl":1772,"updatedAt":1810},"mcore-create-issue","investigate CI failures and create issues","Investigate a failing GitHub Actions run or job and create a GitHub issue for the failure.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1803,1806,1807],{"name":1804,"slug":1805,"type":15},"Debugging","debugging",{"name":1794,"slug":1795,"type":15},{"name":1808,"slug":1809,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":1812,"name":1812,"fn":1813,"description":1814,"org":1815,"tags":1816,"stars":1771,"repoUrl":1772,"updatedAt":1823},"mcore-linting-and-formatting","lint and format Megatron-LM code","Linting and formatting for Megatron-LM. Covers running autoformat.sh, tools (ruff, black, isort, pylint, mypy), and code style rules.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1817,1820],{"name":1818,"slug":1819,"type":15},"Best Practices","best-practices",{"name":1821,"slug":1822,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":1825,"name":1825,"fn":1826,"description":1827,"org":1828,"tags":1829,"stars":1771,"repoUrl":1772,"updatedAt":1837},"mcore-migrate-gpt-to-hybrid","migrate Megatron-LM models to HybridModel","Migration guide for moving Megatron Core GPTModel checkpoints, model providers, training commands, and layer mappings to HybridModel.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1830,1833,1836],{"name":1831,"slug":1832,"type":15},"Machine Learning","machine-learning",{"name":1834,"slug":1835,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":1839,"name":1839,"fn":1840,"description":1841,"org":1842,"tags":1843,"stars":1771,"repoUrl":1772,"updatedAt":1850},"mcore-onboard-gb200-1node-tests","onboard functional tests for GB200","Onboard 1-node GitHub MR functional tests for GB200 from existing mr-scoped 2-node tests.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1844,1847],{"name":1845,"slug":1846,"type":15},"QA","qa",{"name":1848,"slug":1849,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":1852,"name":1852,"fn":1853,"description":1854,"org":1855,"tags":1856,"stars":1771,"repoUrl":1772,"updatedAt":1861},"mcore-run-on-slurm","launch distributed training jobs on SLURM","How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1857,1858],{"name":13,"slug":14,"type":15},{"name":1859,"slug":1860,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":1863,"name":1863,"fn":1864,"description":1865,"org":1866,"tags":1867,"stars":1771,"repoUrl":1772,"updatedAt":1875},"mcore-split-pr","split pull requests to reduce review load","Split a PR into multiple PRs to reduce the number of required CODEOWNERS reviewer groups.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1868,1871,1872],{"name":1869,"slug":1870,"type":15},"Code Review","code-review",{"name":1794,"slug":1795,"type":15},{"name":1873,"slug":1874,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":1877,"name":1877,"fn":1878,"description":1879,"org":1880,"tags":1881,"stars":1771,"repoUrl":1772,"updatedAt":1884},"mcore-testing","run and manage Megatron-LM tests","Test system for Megatron-LM. Covers test layout, recipe YAML structure, adding and running unit and functional tests, golden values, marker filters, and CI parity.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1882,1883],{"name":1845,"slug":1846,"type":15},{"name":1848,"slug":1849,"type":15},"2026-07-14T05:25:54.928983",{"slug":1886,"name":1886,"fn":1887,"description":1888,"org":1889,"tags":1890,"stars":1771,"repoUrl":1772,"updatedAt":1895},"nightly-sync","manage nightly main-to-dev sync workflows","Domain knowledge for the nightly main-to-dev sync workflow. Covers merge strategy, CI architecture, failure investigation, and known issues.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1891,1894],{"name":1892,"slug":1893,"type":15},"Automation","automation",{"name":1781,"slug":1782,"type":15},"2026-07-30T05:29:03.275638",496,{"items":1898,"total":1994},[1899,1916,1926,1940,1950,1965,1980],{"slug":1900,"name":1900,"fn":1901,"description":1902,"org":1903,"tags":1904,"stars":23,"repoUrl":24,"updatedAt":1915},"accelerated-computing-cudf","accelerate data processing with cuDF","Official NVIDIA-authored guidance for NVIDIA cuDF GPU DataFrames, pandas acceleration, dask-cuDF, ETL, joins, groupby, CSV\u002FParquet I\u002FO, nullable semantics, and multi-GPU DataFrame workloads.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1905,1908,1911,1912],{"name":1906,"slug":1907,"type":15},"Data Analysis","data-analysis",{"name":1909,"slug":1910,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":1913,"slug":1914,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":1917,"name":1917,"fn":1918,"description":1919,"org":1920,"tags":1921,"stars":23,"repoUrl":24,"updatedAt":1925},"aiq-deploy","deploy and manage NVIDIA AI-Q infrastructure","Use when asked to install, deploy, run, validate, troubleshoot, or stop NVIDIA AI-Q Blueprint infrastructure.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1922,1923,1924],{"name":13,"slug":14,"type":15},{"name":1859,"slug":1860,"type":15},{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":1927,"name":1927,"fn":1928,"description":1929,"org":1930,"tags":1931,"stars":23,"repoUrl":24,"updatedAt":1939},"aiq-research","conduct deep research with AI-Q","Use when asked to run deep research or AI-Q research through a reachable NVIDIA AI-Q Blueprint backend.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1932,1935,1936],{"name":1933,"slug":1934,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":1937,"slug":1938,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":1941,"name":1941,"fn":1942,"description":1943,"org":1944,"tags":1945,"stars":23,"repoUrl":24,"updatedAt":1949},"amc-run-sample-calibration","run AMC sample dataset calibration","Run end-to-end calibration on the shipped sample dataset (sdg_08_2_sample_data_010926.zip) against a running AMC microservice. Use when user says 'test sample dataset', 'run sample calibration', 'verify AMC install', or 'launch and test'.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1946,1947,1948],{"name":1906,"slug":1907,"type":15},{"name":9,"slug":8,"type":15},{"name":1848,"slug":1849,"type":15},"2026-07-17T05:29:03.913266",{"slug":1951,"name":1951,"fn":1952,"description":1953,"org":1954,"tags":1955,"stars":23,"repoUrl":24,"updatedAt":1964},"amc-run-video-calibration","calibrate video datasets with AutoMagicCalib","Calibrate a new dataset from pre-recorded video files via the AutoMagicCalib REST API. Use when user has local MP4s and says 'calibrate my videos', 'run AMC on these videos', or similar. For RTSP\u002Flive streams, use amc-run-rtsp-calibration instead.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1956,1957,1960,1961],{"name":1892,"slug":1893,"type":15},{"name":1958,"slug":1959,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":1962,"slug":1963,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":1966,"name":1966,"fn":1967,"description":1968,"org":1969,"tags":1970,"stars":23,"repoUrl":24,"updatedAt":1979},"amc-setup-calibration-stack","deploy AutoMagicCalib microservice with Docker","Launch AutoMagicCalib microservice and web UI from NGC release images via Docker Compose. Use when user says 'deploy auto calibration', 'launch auto calibration', 'launch AMC', 'start MS+UI', or 'set up auto-magic-calib'. Requires NGC API key.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1971,1972,1975,1976],{"name":13,"slug":14,"type":15},{"name":1973,"slug":1974,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":1977,"slug":1978,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":1981,"name":1981,"fn":1982,"description":1983,"org":1984,"tags":1985,"stars":23,"repoUrl":24,"updatedAt":1993},"cudaq-guide","develop quantum applications with CUDA-Q","CUDA-Q onboarding guide for installation, test programs, GPU simulation, QPU hardware, and quantum applications.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1986,1987,1990],{"name":9,"slug":8,"type":15},{"name":1988,"slug":1989,"type":15},"Quantum Computing","quantum-computing",{"name":1991,"slug":1992,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305]